Skip to content
View Coff0xc's full-sized avatar
🎯
Focusing
🎯
Focusing
  • Security
  • Hong kong

Highlights

  • Pro

Block or report Coff0xc

Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Coff0xc/README.md

Typing SVG


GitHub followers  GitHub stars  Profile views


$ whoami
Coff0xc - AI security researcher.
专注 AI 红队、LLM 应用安全、Agent 运行时滥用、MCP/工具链风险。

$ mission
Build reproducible tools for authorized testing and defensive validation.
构建可复现的授权测试工具,把攻击路径转化为防御证据。

$ current-focus
Prompt injection -> Agent hijacking -> MCP exploitation -> Tool poisoning
提示词注入 -> Agent 劫持 -> MCP 利用 -> 工具投毒

Research Areas / 研究方向

Area / 方向 What I work on / 关注内容
LLM and agent red teaming / 大模型与 Agent 红队 Reproducible attack chains for authorized model, agent, and workflow assessment. 授权场景下的大模型、Agent、工作流攻击链复现与评估。
MCP and tool security / MCP 与工具安全 Tool poisoning, unsafe tool invocation, sandbox boundaries, and protocol-level abuse paths. 工具投毒、不安全调用、沙箱边界与协议级滥用路径。
Secret and token exposure / 密钥与 Token 暴露 GitHub/API key discovery workflows, validation safety, and remediation-oriented evidence. 面向修复的密钥发现、验证安全和证据产出。
Defensive automation / 防御自动化 CI-friendly reports, SARIF output, ATT&CK-style mapping, and repeatable test artifacts. CI 报告、SARIF、ATT&CK 映射和可重复测试产物。

Featured Projects / 代表项目

Project / 项目 Focus / 重点 Defensive value / 防御价值
AutoRedTeam-Orchestrator AI red team orchestration with SDK, CLI, MCP, Docker sandboxing, and CI artifacts. AI 红队编排、SDK/CLI/MCP、Docker 沙箱与 CI 证据。 Turns ad hoc testing into repeatable assessment workflows. 把零散测试沉淀为可复现评估流程。
catchclaw Multi-platform agent security testing with DAG-style attack chains and Rust components. 多平台 Agent 安全测试、DAG 攻击链与 Rust 组件。 Helps evaluate agent behavior, tool trust boundaries, and exploit paths. 评估 Agent 行为、工具信任边界和利用路径。
LLM-Security-Assessment-Framework LLM assessment framework covering jailbreak, prompt injection, and adapter-driven testing. 覆盖越狱、提示词注入和多适配器测试的大模型评估框架。 Gives teams structured scenarios for measuring model and app resilience. 为模型和应用韧性测试提供结构化场景。
CTF-MCP MCP server for CTF workflows across web, crypto, pwn, and challenge automation. 面向 Web/Crypto/Pwn/自动解题流程的 CTF MCP Server。 Demonstrates how tool-rich agents can be constrained, tested, and audited. 展示富工具 Agent 如何约束、测试和审计。
Github-API-scan GitHub API key scanning and validation workflows for multiple AI providers. 多 AI Provider 的 GitHub API Key 扫描与验证流程。 Supports secret discovery, exposure triage, and remediation evidence. 支持密钥发现、暴露分级和修复证据。
coffee-skill Codex/agent workflow skill pack for dev, RAG, API/data, and security review tasks. 面向开发、RAG、API/数据和安全审查的 Agent 工作流技能包。 Encodes repeatable engineering and security workflows for agentic development. 固化可复用的工程与安全工作流。

Operating Principles / 行动原则

  • Authorized testing only / 仅用于授权测试:research code is intended for owned systems, lab environments, CTFs, or explicitly approved assessments.
  • Evidence over claims / 证据优先:prefer reproducible commands, logs, SARIF, screenshots, and minimal proof artifacts.
  • Defender-first release style / 防御者优先:publish methods in a way that helps detection, hardening, and regression testing.
  • Small tools, sharp boundaries / 小工具、清边界:avoid hidden production assumptions, credential leakage, and irreversible actions.

Stack / 技术栈

Python, Rust, Docker, Linux, Git, GitHub, VS Code, Bash

MCP Protocol LLM Security Agent Security Prompt Injection Tool Security


Activity / 活动

Private contribution visibility is controlled by GitHub profile settings. This README uses the contribution graph generated from the visible profile activity instead of third-party stats cards.

私有贡献展示由 GitHub 个人资料设置控制。 这里使用基于可见 Profile 活动生成的贡献图,不使用容易漏私有数据的第三方统计卡。

GitHub profile activity

Contributions / 贡献

Contribution graph animation
COFF0XC :: break carefully, document clearly
谨慎验证,清晰记录,让防御更可复现

Pinned Loading

  1. AutoRedTeam-Orchestrator AutoRedTeam-Orchestrator Public

    MCP-native security automation workbench (SDK + CLI + MCP) — authorized testing + static AI/MCP attack-surface self-audit to SARIF / GitHub Code Scanning

    Python 260 53

  2. Github-API-scan Github-API-scan Public

    ���业级AI密钥管理平台 | 23平台全覆盖 | 深度验证(余额/权限/评分) | 实时WebSocket推送 | Web可视化Dashboard | Docker一键部署 | 生产就绪

    Python 79 40

  3. catchclaw catchclaw Public

    CatchClaw v5.3.0 — 多平台 AI Agent 安全评估工具,支持 OpenClaw / Dify / FastGPT / LobeChat / NextChat / AnythingLLM / Flowise / RagFlow 等 9 大平台。78 条 DAG 攻击链 | 78 个 Exploit 模块 | ATT&CK 阶段映射 | 多平台指纹识别 | Async To…

    Rust 31 5

  4. LLM-Security-Assessment-Framework LLM-Security-Assessment-Framework Public

    FORGEDAN — An Evolutionary Framework for LLM Security Assessment | 6 Attack Methods | 18 Model Adapters | Vue 3 Dashboard | Web Security Scanning

    Python 24 6

  5. coffee-skill coffee-skill Public

    Installable Codex/AgentSkills workflow pack for dev, Agent/RAG, API/data, Office artifacts, research diagrams, and authorized security review.

    Python 16

  6. vero vero Public template

    Evidence-Driven Autonomous Red Team Agent - The first AI penetration testing tool you can actually trust

    Go 6 1